Developments in post-marketing comparative effectiveness research

Abstract

Physicians and insurers need to weigh the effectiveness of new drugs against existing therapeutics in routine care to make decisions about treatment and formularies. Because Food and Drug Administration (FDA) approval of most new drugs requires demonstrating efficacy and safety against placebo, there is limited interest by manufacturers in conducting such head-to-head trials. Comparative effectiveness research seeks to provide head-to-head comparisons of treatment outcomes in routine care. Health-care utilization databases record drug use and selected health outcomes for large populations in a timely way and reflect routine care, and therefore may be the preferred data source for comparative effectiveness research. Confounding caused by selective prescribing based on indication, severity, and prognosis threatens the validity of non-randomized database studies that often have limited details on clinical information. Several recent developments may bring the field closer to acceptable validity, including approaches that exploit the concepts of proxy variables using high-dimensional propensity scores, within-patient variation of drug exposure using crossover designs, and between-provider variation in prescribing preference using instrumental variable (IV) analyses.

Figure 1

Causal diagrams demonstrating the mechanics of confounding and three approaches to reduce confounding by unmeasured factors.

Figure 2

Intended and unintended treatment effects and the potential for confounding by indication.

Figure 3

Drug utilization patterns guide the choice of non-randomized study designs.

Figure 4

Regions of non-overlap of the exposure PS distributions of two treatment groups. In this example, study patients were restricted to those with largely overlapping exposure PSs by trimming patients with extreme PS values.

Figure 5

Restrictions to study populations typically applied in comparative effectiveness research. Modified after Schneeweiss et al.

Figure 6

Physician-prescribing preference. In this example, physicians treating study patients were separated into those who have either a strong preference to prescribe drug A (on the right side) or a strong preference not to prescribe drug A.

Figure 7

Sensitivity analysis of residual confounding. This example by Psaty et al. evaluates the effect of unmeasured confounders on the association between calcium channel blocker (CCB) use and acute MI (apparent relative risk or ARR = 1.57). The study assumed a prevalence of the unobserved confounder (P_C) of 0.2 and a prevalence of CCB treatment (P_E) of 0.01. Each line splits the area into two: the upper right area represents all parameter combinations of the association between confounders and drug use (OR_EC) and the strength of the association between the confounder and outcome (RR_CD) that would create confounding by an unmeasured factor strong enough to move the point estimate of the ARR (ARR 1.57) to the null (ARR 1) or even lower, i.e., make the association go away. Conversely, the area to the lower left represents all parameter combinations that would not be able to move the ARR to the null.